Longevity of DVD-R and CD-R (Rewrite from original poster)

[bristol22 at softhome.net]

I am the scoundrel that made the original post on this
topic that opened a flood of discussion. Members brought
up details that I never addressed; and I learned from
everyone's posts. Thank you! This may be useless, beating
a dead horse, and off topic.  I am re-submitting here my original
6 point comment with corrections and links, and with ADDITIONS
as points: 7, 8, 9, 10, 11, and 12 to cover other areas people
have brought up, etc.
  Richard Bristol 

 ----------------------------------------------------------
Archival longevity of DVD and CD media may be a good topic
for the computer preservationist. I have studied this quite a bit.
My first conclusion, including from crude accelerated life
and torture tests I made myself, is that the "100 year" claims
some people made are likely baloney. A research article appeared
in the German version of Q't magazine (unless I miss-spelled
that), where they did some tests on raw error rate on just
burned DVD+R media, using some $50,000 equipment. None of the media
they tested had a low enough raw error rate to meet the requirements
of the DVD+R standard. Yet they sell it. (Mitsubishi did the best.)
Then there are many hobbiests looking at raw error rates on DVD
media using Lite-On drives with firmware than can report on
raw errors. For example, hobbiests measure and post on sites like:
http://www.cdrinfo.com/Sections/Reviews/Print.aspx?ArticleId=11657
My overall conclusion is that most DVD media on the
market sold by no-names, and by the names Sony, Maxell, Memorex,
3M Imation, etc is not very good. These companies re-badge
whaterver is cheapest this month from whatever factory. My
least favorite factory is CMC Magnetics Corporation in Europe.
Next worst in my view are some Korean and Taiwanese factories
(K-Well, In Young, and Lead Data=Great Quality=GQ). TDK seems to
be the only advertised brand that makes their own DVD-R that you
can actually buy in a USA retail store (although one member said
maybe they don't make it themselves), unless the store carries Verbatim.
I just verified that Pioneer and Verbatim buy their media from the
good Japanese companies, Mitsubishi (MCC), Taiyo Yuden (TY), and Mitsui.
Anyway, Pioneer and Verbatim consistently have low raw error rates.
A low error rate when freshly burned gives greater margin that
as the errors increase as the media ages, error correction can
still take care of it. Also, these companies have better engineering
about the life span of the organic dyes and such used in the
data layer. So when I want it to last, I do this:
1) Use a good burner with low raw error rate after burn/read and
 good media-to-other-machine cross compatibility. I used to
 recommend the Pioneer DVR-105 and below, but the DVR-107 and
 DVR-108 are poor at this. Now I recommend BenQ DW1620 ($52)
 DVD+-RW double layer drive. Most burners have new firmware you
 can flash even if you bought it yesterday. A lot of the effort
 in the new firmware is decreasing the error rate on some media,
 so it is often worth flashing it.
2) My first choice of media is either Pioneer or Verbatim DVD-R,
 followed by DVD+R from the same two makers.  Third choice brand
 is TDK. I don't trust much else. Double sided and double layer
 DVD are a little imature at this point.
3) The freeware Windoze program DVDinfo.exe will tell you who made
 your mysterious media (reads it off the media). E.g. MCC001 is
 made by Mitsubishi.
4) All RW media (DVD-RW DVD+RW CD-RW) have poor archival life.
 Think about it: with RW, instead of burning a pit in the data
 layer, you are fooling around with glossy or matte finish
 depending on how quickly a melted liquid re-freezes. Official
 tests, and my own tests, show poor life. A little sunlight-UV
 can erase it.
5) I don't burn more than 85% of the capacity of a DVD. DVDs are
 a sandwich, lexan (polycarbonate) on both sides, data layer
 chemistry in the middle. The spiral starts on the inside (opposite
 LP records). If you leave 3/4" unused at the outside edge, it
 will take longer before the Ozone and 02 and other environmental
 exposures that attack the data chemistry at the edges of the
 lexan sandwitch actually reach your data. So on a "4.7GB" disk
 which in real life could hold 4.3GB, I burn 3.8GB.
6) DVD-R or DVD+R is better than CD-R, because it is a sandwich.
 DVD also has stronger error correction. CD-R often dies from
 the label side, by getting bumped against the drive tray,
 the jewel case, a pen...  The paint layer that protects your data on
 the label side is very thin and fragile. CD-R also die from chemical
 errosion from ozone, fingers, marking pens, etc on the label side.
 I write on them only in the center no-data area. DVDs, being
 a sandwich, don't have those problems.
7) If you really care about it's archival life, hedge your bets
 by burning two copies, on DIFFERENT media. Select two good
 media types made by different companies. Each may have, just
 a guess, a 10% chance of self-destructing within 20 years.
 This would be because of a design or implimentation mistake.
 For example perhaps the sandwich adheasive includes a chemical
 that eventually reacts badly with the data, or some impurity
 in manufacturing at that time. If the two media are very
 different, their probabilty of failure approaches statistical
 indpendence; and if independent then the chance that both are
 unreadable would approach 0.1 * 0.1 or only a 1% chance.
8) Because the data is sandwiched in plastic, my belief is that
 single-layer single-sided DVD-R or DVD+R as a class will last longer
 then CD-R as a class. However, some applications require CD-R.
 For archival CD-R I have used extensively and can recommend
 "BASF by EMTEC" (Emtec in USA) "Ceram Guard" CD-R BASF CD-R74
 Maxima Ceram DA (really made by Taiyo Yuden). (Beware some
 different Emtec/Basf CD-R are made by CMC and are poor.)
 However "Ceram Guard" is truely hard to buy in the USA. I bought
 mine for $2 each from the retail chain "Guitar Center" in person
 (guitarcenter.com). They were marked "for audio use" but that does
 not matter. They are NOT gold. Their feature is that a ceramic
 coating is vacumn sputter-coated on the labeled. You could probably
 use a ball point pen on it. They are by a mile the most sturdy
 CD-R I have ever seen. Unlike most CD-R, taps to the label side
 don't kill them. Out of about 1000 CD-R I have seen or used "in the
 field", I have seen about 25 dead CD-R's (that were once good)
 And 22 of those 25 were at least slightly abused. (Sun, moisture,
 tap on drive tray on label side when inserting, or tap on jewel case)
 Only 3 of 25 appeared to die from purely with "from the inside bit rot".
 Of those 3, 2 were made by one company: GQ Great Quality, Taiwan. So
 I am mostly interested in the label side coating.
 These ceramic coated disks are mentioned here:
   http://www.cdmediaworld.com/hardware/cdrom/basf.shtml
 BASF/Emtec also sells these as "archival" but I would want to know
 who made it for them before buying. They are real gold:
   BASF/Emtec CD-R Gold Digital Photo
   BASF/Emtec DVD+R Digital Video Gold
 Occasionally you can find "Kodak InfoGuard". They are expensive. They
 have a very tough label side, and use gold. They are semi-discontinued,
 and were sold to medical and business markets. If you have a Kodak
 "Picture CD" from film processing you may have one there.
   Kodak digital science InfoGuard, maker: Kodak Japan Limited,
     Phthalocyanine 1X - 4X
   Kodak Gold Ultima InfoGuard, maker: Kodak Japan Limited,
     Phthalocyanine 1X - 4X
 More practically speaking, you CAN actually buy Mitsui "MAM Gold CD-R"
 for about $1 each now. They use real gold, and have a decent
 paint layer they call "diamond coat". (But it's not anywhere
 near the toughness of "Ceram Guard" or even "InfoGuard".)
 Described for sale here:
 http://store.mam-a-store.com/standard---archive-gold.html
9) I don't know if high-spin readers contribute to media death. But
 too high a speed definitely contributes to "can't read Table of
 Contents" and "ECC Error" on read, and marginal quality burns.
 Plextor CD-R burners come with PlexTools(?) software specifically
 to slow them down. You can sometimes read a CD-R this way that
 otherwise can't be read on newer faster readers.  To read a
 CD-R whose errors are too high for most all readers, I prefer
 these to read it: Plextor 4x CD-RW drive, or 12x (not 8x).
 Plextor 20x and above are not as good. Cheap and I have a
 few: Panasonic 12x just plain reader. Sometimes other 4x readers
 by NEC, Panasonic, Pioneer, Sony.
10) I like to try, when burning, for the lowest raw-error-rate (that
 some machines can tell you when you read). Published tests show
 excessive speed hurts. My rule of thumb when burning is:
 Never exceed the lesser of the media's rated speed, or 2/3 of
 the burner's maximum speed. So if a CD-R burner will burn at 48x,
 and the media says it is 80x, I will burn it at around 12x or 24x.
 There is such a thing as too slow also, for a particular burner.
 The sweet spot gives you a lower raw error rate, and greater
 cross-compatibility when reading it on many different drives.
11) A disk that takes a long time to "settle" (usually light goes
 out) on insertion is a possible indication of a high error rate
 and likelyhood of complete table of contents failure later.
12) The free Windoze program md5summer.exe will generate a text
 file with .md5 ending for any directory tree of files. Each line
 in the file is the md5, a space, *, and then the pathname;
 one line per file. I put one of these .md5 files in the top
 level of my CD-R or DVD-R burns.  Then, in Windoze a couple years
 later, double click that .md5 file and it does the
 opposite, and confirms the md5 of each file on the media.
 I have seen a few CD-Rs that were damaged, that in Windoze gave no
 error message at all when read, but the data read back was
 actually wrong. I use md5summer also before ftp etc. transfers.
 Will someone better informed say how this is best done and
 commonly done in linux and unix? Perhaps
   find .  -exec md5sum '{}' \; > md5sum_list
 or something? And the checking later? 

 Richard Bristol  (the same guy who recently asked for
   help reading 9-track tar tapes)